Watch our original tear-down of the 5700 XT here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-BpctR5TFJjk.html And our review here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE--SAWtKEIYbw.html And find the GN toolkit here: store.gamersnexus.net/products/gamersnexus-tear-down-toolkit
unrelated to the cooling issue, but why does the 5700 card even exist. The 5700xt is only fifty bucks more and competes with the 2070. Why not just save up that little bit more and buy the xt?
ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-2hIOW6qRbKM.html Can you look into this "pad", it's supposedly a polymer gel? that sets into a firmness that resembles a pad.
Hey Steve, would you mind doing an aftermarket cooler like the accelero xtreme iv or raijintek morpheus ii. The idea being to test the raw performance of the 2070 super versus 5700xt when the cooler isn't a factor. Waterblock would also work, of course, so just whatever is simplest or on hand.
Reference cards are only good for water blocks. (Except for fan models, like NVidia went with.) Those blowers just don't work well.. even with good mounting for the heatsink.
@@qzcc5550 Well I consider that part of the cooler package, but point taken. There may be better cooler designs, but going with a thermal pad over a good paste is probably the big core issue.
I found that adding thicker rubber pads on bracket in the area that pushes down on the back of the PCB was much more effective than the washer method at increasing the mounting pressure between the die and the cold plate. You also wont be able to see the mod after the fact. I used thick foam double sided tape, and added layers until i got the optimum contact pressure; used that contact paper idea you guys introduced. I used this method on my Radeon VII, the Washer method is rather sketchy in my book, and looks it to.
@@GamersNexus The pressure from the rubber fatigues over time I have seen hacks like that before on say.. trains.. But there is springs for compact installations you can look at. Lesjofors got some neat ones.
off topic but back in the red ring of death days for the x360 after you baked the chip that came desoldered I would use plastic thin washers under the mounting hardware as well as flex the bracket a bit to get much better mounting pressure. This worked out pretty good and I did a lot of 360's this way and all worked fine for a very long time after except one of them which had gotten another problem not related to the red ring of death. Because these were pretty much classified as dead there was little risk of breaking anything because it was already non functional.
@@GamersNexus You guys might want to try that method too. Also after that some lapping if you dont mind :) On the paste reccomendations you've made. I think TG kryonaut ''breaks'', after 80C so It would be a good choice for this card. Know that from Vega 56, Vega 64 owners.
Dr. Lisa Su: I want good thermals on Navi. AMD Engineer: The technology actually doesn't exist. Dr. Lisa Su: STEPHEN BURKE WAS ABLE TO BUILD THIS IN A CAVE! WITH A BOX OF SCRAPS! AMD Engineer: I'm sorry. I'm not Stephen Burke.
Ok, I've done this mod also like another chap in the comments. I used thermal grisly kryonaut, 1mm thick silicone thermal pads and some plastic washers. With the washers, I made sure the hole was big enough for the small post that pokes through the pcb to fit within, and also kept the springs on. So order is screw, bracket, washer, spring. The spring also fits inside the washer. I also put 4 1mm thick strips of plasticard under the four rubber feet around the center of the mounting bracket (1 strip on each rubber foot), to increase the pressure at the center. Then did an auto undervolt in wattman and now I site at 2020mhz over prolonged furmark 1080p runs, getting to a junction temp of about 86 after saturation. Clock may dip a bit on saturation, but still above 1900.
AMD: We invested millions in bringing you these new cards GN: we spent bout 4 cents and made it better :p Doesnt that pressure test paper look like the GN logo??
Dumd Amd wasting opportunity again , should have made 2fan design and GPUs could have clocked higher and match rtx2070super performance for less money.
I used washers on the heatsink assembly for laptops through the years,because the springs were a terrible rate. I thought I was alone in my thinking and method,but now I realize I'm not alone.
3:26 - You NEED to use PLASTIC washers. You can see in the closeup shot at this time index that the metal washer is touching some resistors on the back side of the card. This could lead to a short and failure once the coating wears off.
I used metal washers and there is no problem for now. But my washers were not as big as the ones in this video actually they had the perfect size and fit perfectly under the metal thing
When I used to do reflow repairs of GPUs I found that using a copper shim with good quality thermal paste like arctic silver worked best. The trick is to test fit different thickness copper shims and to have some resistance to moving, when the heatsink is reassembled take a pick or small screwdriver and try to move the shim. You want enough resistance that the shim will move when pressure is applied. After finding the right ship apply the thermal paste and it’s good to go. You don’t want a ton of pressure on the GPU die as pushing the chip down too much will cause damage. Done correctly the shim has enough room to move around from thermal expansion and provides better cooling. You can find a copper shim assortment on eBay and they are cheap. On average I would get at least a 15c - 20c degree decrease in temperature.
For anyone interested, the paper is called Fuji Impression paper. It's cheap to buy, but it costs around $50 to analyze a single piece unless you buy their scanner which costs $10k. The more red does mean better contact, but it doesn't mean Forces are evenly distributed. The Red color needs to be scanned and measured in pixels to understand its pigment and its saturation, then you can call it good or bad pressure(force) distribution.
if you need that info so much, just Google any other cooler mounting instruction with cross plate on the back and add to that some washers between gpu board and cross plate that's it... so basic, ofc he wouldn't waste his time and resource on rendering another minutes of video, it's a basic routine. Tons of tutorials how to apply all that over internet. don't be lazy, or if you are too afraid to do it, just let someone experienced do It fo u
@@joeallan3706 and for extra min 50 dollars its even more in my country so ı will buy blower but ı dont think it will be a problem for me ı play max 1 or 1.5 hours and rest for another 30 min
Checkt out pcgh.de. It's PC Games Hardware, a German magazine. Am on a mobile connection so can't include the link, but they had a video yesterday. Still interesting even when you don't understand him talking. Considering you can take these coolers from gen to gen and always go for the reference card it's very compelling.
Try being in australia.. 0.20c for the washers, plus import tax, plus customs tax, plus wholesaler markup, plus retailer markup, plus GST, grand total of $89.72USD, plus exchange rate conversion is a grand total of $146.56AUD
Yeah, don't do this. Must use plastic, or those old school red paper washers under metal. Another problem is metal can penetrate the PCB coating and contact the copper traces.
Those HM03 pads are good, but not for high density, high TDP parts. AMD should focus on increasing the mounting pressure of their coolers and start using high conductivity thermal pastes instead. Preferably a trilinear ceramic based compound to reduce the effects of ageing.
I understood the use of the pad in Radeon 7,as it was a big die with 4 HBM stacks, but I don't think this could be a problem for the radeon 5700 when the die is a lot smaller, unless the heatsink has really really bad tolerances.
@@hantzleyaudate7697 And so? Any partner card uses thermal paste, this was done on r7 cause the cooler was uneven, smaller die means smaller cooler contact so it should be less of a problem.
i think they started using it with r7, then seeing some sort of production side efficiency of using pads vs. paste they decided to use it for all cards. the only way i can see them doing it with negative thermal consequences is if it is saving them money somehow.
@UCH2l_lS7uiIF5Vffbq0cnFQ TBF the video title allows you to deduce the findings of the video. Stuff like this happens because companies rush products to market. If there is another explanation you think is likely then do share.
@@tonybarden8563 True. But let's also consider that AMD's thermal solutions for their GPUs are notoriously garbage. To compete they have to take it seriously. It's not good enough to produce a product that competes on price and perf but then falls flat on its face because it's noisy and hot.
@@downwiththatsortofthing624 You missed out your part; triggered keyboard warrior. A perfect trifecta! Not entirely sure what you're standing up for though. Now that we've all had time to watch the video (which I have done - you'll find soothing to hear) do you propose we alter our hideously premature conclusions?
@@stephenwakeman3074 well the blower is a 'fit-all' solution ,but then you do have to ask at this sort of price ,most buyers are actually pc clued up with decent case venting etc ,so a blower for the masses may not be a good idea ,but hey how on earth would a pc part company know there market is for a gamer ....lol i wasnt defending the blower at all mate ,just pointing out the business/automation process would be less good at doing the ideal job as a tech channel by hand
for everyone complaining about AMDs design, consider that they can't just add 1mm washers to all cards in production. The amount of spring force will depend on the thickness tolerances/variances of the washers, the gpu die, bend angle of the spring etc.. maybe a total of +-1mm of total tolerances. The spring is designed to not add too much pressure at the max, and "enough" at the minimum. Increasing the spring stroke may get you into a pressure that the VC is not designed for. But this is not a problem when you check with pressure paper and adjust the washer thickness accordingly. Which would add alot on the pricetag in mass production. Steves comment at the end hits the nail on the head.
Thanks for the video. I applied this mod using nylon #4 size washers and my 5700 XT is now running around 78c with the fan at 2100 rpm. The card maintains about 1900MHz. Again thanks for the affordable mod for cooler temps and little to no throttling. :)
The spring effect of the bracket can be increased if you bend all 4 legs with 2 pliers. So it needs much more force to push it down and you also avoid bending the pcb this way.
there is one more concern with the metal washer. its that if it was not installed correctly, it could cause a short circuit, or even when it is being installed corrected, if something went wrong like the washer scratched the surface of the PCB and exposing data lines and causing a short circuit, the damage could be lethal to the card. to me, plaster is a much safer choice and more reasonable choice.
Also the grille in the back mounting bracket restricts airflow causing the fan to run in a semi stalled state. Lay the case on its side so you can run the card with the plate removed as a test. You will see lower temperatures and reduced noise. Modding the back mounting bracket with saw and file's, to maximise the opening size is pretty easy.
I hope the right AMD people are watching this and manage to get these improvements into their Radeon cards. The washers won't be a huge added cost/effort and the positive consumer feedback (especially regarding to temp issues) will be worth it.
Ok, first of all I have been PC gaming and building rigs for 20 years but I have never took apart a graphics card before. I was nervous to do it but I got the plastic washers, some Kryonaut and some thermal pads (for the back plate) and did it. Now, even with the stock fan curve it is not going above 79C on unigine benchmark/stress test. Thanks so much for making this video and the tear down one, it was a big help.
@3:33 looks like the washer on the left is over the top of those 0402 resistors. Small surface mount components like that are susceptible to cracking and fracturing which alters their properties or causes them to go open circuit completely. It's probably a better idea to find a smaller diameter washer or a small length of tube.
40db measured at an arbitrary 20" and on an open test bench. So install in a case and leave the cover on and you will barley hear the card whoosh, run the latest drivers for the fan curve and of course run games and not Torture Tests so temps will be normal and no throttling will be observed.
So I bios flashed my 5700 reference card to the XT bios as soon as I got the card, went from 78°C max temp in Modern Warfare at max settings @ 1080p, up to roughly 86°C after the flash. Today I applied thermal paste, lightly sanded the contact plate on the vapor chamber with 2,000 grit paper, and made my own washers with a guitar pick and a drill, and I'm back down to an average max of 80°C. But I went from average 100fps in-game to capping at 144fps. Overall pretty worthwhile, but I will be getting a waterblock for it soon
You should sell this a kit on your website - it took a lot of tech savvy people to make a good product but it takes a tech genius to take that good product and make it great.
@@thenetpagan yeah but mounting and spacing is the same again and the GPU generally sits in the same spot on the PCB on every reference full length card.
I date back to a time where if a GPU even had a fan it was a single 80x80 glued to the board and if if failed it was time for an upgrade anyway. I say that to buy some sympathy when I admit that I never knew about much less checked on these issues with ny 5700 XT from early 2020. It was at least amusing given their was a happy ending. Between the shit VRAM mounting to the pads and the passage of time, if one were to try to the “penny” trick, they’d have needed to stand said pennies on their side to bridge the gap between the VRAM and the thermal pads. I couldn’t confirm what version I had and scree types seemed to vary randomly anyway, so I went my own way and just used one less washer than it took before any given screw could not reach the threads, which was often less than videos like this recommend (making me assume I have a refresh model with shorter screws) and yet I covered a near half inch gap, and that’s with a bamboo stick stuck through the bottom of the case so as to support the end of the card from sagging. I honestly don’t know how that card could separate that badly, but holy fuck am I glad I noticed it now (and thanks to thermal throttling because every card I had prior that failed did so by exploding) because it runs better than new.
@Alex Davies I've seen other channels test 1440p + ris vs true 4k and say the quality was very similar. So even though they can have different uses you would test bfV / ffxv / whatever else has dlss with both features (1440p ris vs "4k" dlss vs true 4k on both) with 5700xt and rtx 2070 super. This could show if either could do a "4k like" experience at high frames without spending 2080ti bucks.
It's no different than an upscaler on your tv. X Y Tap scale with bilinear or trilinear (depending on noise function) and then apply a sharp filter matrix to harden edges. It can be processed as the buffer flip is called.
The slip drive of "modern" battery driven cordless electric screwdrivers - the thing where you can turn an outer wheel to some number - is actually a newton meter torque dial, which as a thing has been around in the industry (manufacturing) for decades (60's or earlier) before it became "common place" in consumer ware. This is also where all these screw heads come from, Phillips screw heads versus simple slot and Torx versus Phillips, etc. Also you have might have noticed this little dent in one of the corners of a screw head - usually very prominent in Phillips screw heads - that actually is for the mechanized manufacturing, so the machine knows how many turns it did (and initial alignment, etc.). So, no; there is no excuse from a mechanized stand point to not properly fasten those screws. Actually it should be more accurate and repeatable than anything a human does on the home workbench.
Could yall do a video adding a custom triple fan cooler and checking the different in thermals!?!?! The fan kit is called "ARCTIC Accelero Xtreme IV High-End Graphics Card Cooler"
I can confirm this works with the 50th Anniversary card as well, 30 minutes in Heaven benchmark and max Junction temp reached 109, but averaged 106, and the temp averaged 85, while maxing at 86. Also clock speed averaged 1900Mhz without modifying any settings during the run.
Washers under heatsink legs is a laptop thing, which are usually low pressure mounting from pressure coming only from tensioned arms, worse when it's tripod mounts and multi-die one-piece heatsinks. I'm also thinking the process of heating vapour chambers (soldering heatpipes etc to them) increasing their internal pressure, and their thin walls, makes for a bulging effect that requires lapping the mated surfaces afterwards to get them back to flat.
So I know little too nothing about computers and the current hardware market but I just built my gaming PC for college. While running vermintide 2 on ultra I heard a lot of noise from my fans. Which in it of itself isn't too concerning, but I was paranoid that I might overheat the GPU or CPU while playing, so I brought up the temp for both, the CPU being fine but the GPU hitting 93°C . After watching this video that seems pretty normal, but I'm still unsure of where the danger zone lies for a RX 5700 XT
I hope they take this idea and implement it on newly manufactured cards. One great thing about GN is that you guys give free R&D without the fear of being fired.
AMAZING :-0 didnt one of those metal washers seem to be touching a contact on pcb? Wouldnt it help to put washers on rest of backplate to to benefit memory and mosfets?
This literally brought my junction down 23+C. Thank you! (MSI Evoke OC 5700, red bios mod to 2016mhz core clock and 1800 memory clock with voltage curve correction; thermalright 12.8w/mK pads with kryonaut paste running 63 current and 78-79 junction, down from 100-102 junction)
@@goofball1_134 Yo, I went to Lowe's hardware and bought a pack of tiny gold metal washers. Idk what size, but basically the smallest they had. I used them for all 6 screws on the back of my GPU
Feels like Navi was rushed some as cooler should have been better design. Consumer shouldn’t have to fix there own mistake. While love zen 2 are better and Navi is better. AMD should fix them
You can compare thermal diffusivity which control the rate of diffusion in the material. Reality is that when it comes to thermals, measuring conductivity alone isn't enough because there's an equally important property which is volumetric heat capacity. You want a thinner pad/paste to reduce the heat capacity interference from the interface, in general. You may want to read about dual needle thermal property testing compared to heating plate/cooling plate testing, if you see this anyway.
i did it with, of course, kryonaut and washers. but it looks like the same results. maybe. yes, there was thermal pad and cracks when i disassemble the card. my results after the mod are almost the same i think. 84 celsius and 106Tjmax at 2395rpm. with 100% torture test "msi kombustor" . blower is little bit noisy but it is ok . higher rpms are much louder. one thing is that temp is lowering down much faster. and there is few more screws on the card but thats all. easy mod .
ASK:GN --- Have you talked to @HardwareUnboxed about the difference in thermal performance? He seemed to have the most adequate thermal results, while GN had the most poor sample of all the channels i watched. I know you guys have discussed your pre-release findings in the past, and seeing as you both are some of the most thorough reviewers i'd love to see that behind the scenes conversation trying to figure out the solution. For what it's worth i trust both parties did their due diligence, but simply there was variance in early sample quality.
Thank you for this video. My 5700XT arrived today and had be contemplating returning ever since it shipped. This video puts my mind at ease. P.S. I am one of those idiots that loves the design of this cooler.
WOW! the fact that you have to modify the thermals to get a better performance from a card that just drop, makes me wonder if I should wait to pick one up.
That's only the case with Nvidia cards with GPU boost (basically it scales your clock speed to your temps) with AMD cards you are fine as long as you aren't throttling (although running at 100% fan speed can help you achieve higher overclocks with the increased cooling)
@@sadgardevoir3734 wrong, nvidia uses power and temps. putting a nvidia card to 100% fan speed will do NOTHING for performance, because they are already below throttling temp. simply by changing powerlimits you can archieve +10% perf without doing anything else. navi cards throttle on several parameters not just temp.
Easiest solution: just buy the Arctic Accelero Xtreme III or Arctic Accelero Xtreme IV Rev.2 (ca. 50 $ / €) ... No disturbing noises and far better temperatures are the result!
Very cool fix! I'll try to keep an eye out for partner models later on because while performance is nice especially on the XT model, I am not a fan of the noise and heat of the card.
I'm seriously considering doing the washer/paste mod to my 5700xt Anniversary just to see what it can do, but I'm leery to mess with it and void my warranty on something so new. I got really lucky with my card and its maxing at 81C under torture tests, 2020Mhz clock speed and 75% fan, and doesn't seem loud. I don't even have Wattman enabled and haven't made any changes to the card or its software whatsoever. It's worth noting that my friend has a sapphire pulse model card and I tried it in my system, and my blower style card got better thermals and higher clocks, perhaps because of my mini-itx case? I haven't tried my card in his giant computer to compare.
Vincent Carrière Hi, I should have fully explained my thinking. Since empty solder pads are connected through their traces to a circuit, think about what may happen if the circuit from the empty pad were to be cross connected with the circuit of the SMD under the washer.
The GPU silicon is fine. The whole design around it is atrocious. 95 degrees and thermal shutdown on a menu (per Anthony of LTT). Not to mention the objectively ugly "shipment accident" aesthetic.
They say it is as much an airflow thing as a marketing one. But I'm guessing it's mostly marketing, as it looks different, and the shroud wouldn't have to follow the internals
40db measured at an arbitrary 20" and on an open test bench. So install in a case and leave the cover on and you will barley hear the card whoosh, run the latest drivers for the fan curve and of course run games and not Torture Tests so temps will be normal and no throttling will be observed.
I have installed a Accelero IV on my 5700XT - the maximum Temperatures are about 65° with 40% Fan Speed. Amazing Solution for the problem, but be careful.
10mm? Holy hell i think he said 1 mm man. I could be wrong but i believe he said 1mm and 0.7mm. 10mm is like a quarter of an inch that is absolutely massive and would wreck your card for sure
@@GamersNexus can confirm. Have a full tube of Corsair thermal paste. They sell it at the local best buy and is actually pretty good. Got lower temps using it when I replaced my noctua paste after 4ish years so I figured id keep a spare on hand.
Honestly if you guys released a screwdriver kit similar to ifixit's i would immediately buy one. I feel like the full sized screwdrivers in your current kit currently limits the usability for general teardown usage (e.g. security torx bits/t5 and whatnot). If anything with interchangeable bits were released I'd be first in line. Love you guys though❤
How is it even possible to have these thermal issues which can be improved so much by just using washers in the first place? Doesn't AMD test their coolers at all? Why use Blower Coolers? I know that they have their uses in cramped/small Cases, since they blow out the hot air directly outside, but come on that is so niche. The vast majority of Gamers don't use super small ITX Cases.
You know all those fancy bling cases with the tempered glass fronts and sides? Yeah those have worse airflow than most ITX cases.. they are for those too.
@@bdhale34 I have a friend who shall remain nameless. Lets call him Mybrother. Mybrother had a case very similar to the one you describe and did NOT have a blower-style card. It baked its self to death. Sample size of one, I know, but he would have likely had better performance with a blower card that didn't suck on its own heat. The story has a happy ending though: I convinced him to pull the front panel off the case and his temps dropped massively.
So basically vapor chambers are terrible because you always get reduced mounting pressure in the very areas the liquid can make contact with the cold plate, while the places where internal copper reinforcements allow for greater pressure and better heat transfer also preclude said heat to be directly removed by evaporating liquid. The way this seems to be influencing contact pressure could lead to minor changes in vapor chamber geometry, say a production cost reduction by the manufacturer, leading to major changes in local heat load across the die, which I guess may be one reason for AMD implementing dozens of temperature sensors for each GPU and basing their turbo modes in part on any one hot spot, regardless of position in the design.